Research on land cover classification methods for each summer and winter based on the Google Earth Engine platform

This study utilized Landsat remote sensing image data and leveraged the big data processing and analysis capabilities of the Google Earth Engine (GEE) platform． By employing the Random Forest algorithm, this study conducted remote sensing monitoring of land cover changes in Xinjiang during the summer and winter seasons from 2000 to 2022． The findings of the study are as follows: 1) The overall accuracy of land cover classification in Xinjiang during the summer and winter seasons is 95.59% and 91.32%, respectively,with Kappa coefficients of 94.57% and 88.36%． 2) Over the past 23 years, the average annual growth rate of summer cropland in Xinjiang is 0.035%, urban area is 0.005%, bare land area decreases by 0.64% annually, grassland decreases by 0.34% annually, water bodies show an overall annual growth rate of 0.067%, and snow-covered area increases from 2.04% to 2.07%． Shrubland and wetland remain relatively stable． 3) Major changes in summer land cover in Xinjiang occur in the transitions between cropland, bare land, grassland, and artificial surfaces, while winter changes primarily involve transitions between snow-covered areas and other land cover types． Winter snow-covered area exhibits a decreasing trend, while other land cover types show an increasing trend． Snow-covered area decreases from 29.71% to 25.79% of Xinjiang’s total area, grassland increases by 1.47%, bare land slightly expands, and cropland experiences a slight growth of 0.62%． The seasonal products provided by this study for both summer and winter seasons contribute to a comprehensive understanding of the seasonal dynamic characteristics of land cover changes in Xinjiang． This information can be valuable for agricultural production, land policy management, and resource management by offering data support．